The present invention relates generally to a conversational agent capability assessment method, and more particularly, but not by way of limitation, to a system, method, and computer program product for assessing performance, personality and cognitive traits of conversational agents based on natural written language by other conversational agents.
Testing a conversational agent is not a trivial task. A conversational agent has many aspects to be tested from the performance across multiple users to matching the best response for a given answer. Further, conversational agents also may interact amongst each other (i.e., a hotel room booking agent may interact with a rental car booking agent or a first version (V1) of a conversational agent with a second version (V2) of the conversational agent) which adds an additional layer of difficulty for testing.